Prospecting Lighting Applications with Ligand Field Tools and
Density Functional Theory: A First-Principles Account of the 4f
7
−
4f
6
5d
1
Luminescence of CsMgBr
3
:Eu
2+
Harry Ramanantoanina,
*
,†Fanica Cimpoesu,
*
,‡Christian Göttel,
†Mohammed Sahnoun,
§Benjamin Herden,
†Markus Suta,
∥Claudia Wickleder,
∥Werner Urland,
*
,†and Claude Daul
*
,† †Department of Chemistry, University of Fribourg, Chemin du Musée 9, 1700 Fribourg, Switzerland‡Institute of Physical Chemistry, Splaiul Independentei 202, Bucharest 060021, Romania
§Laboratoire de physique de la matière et modélisation mathématique LPQ3M, Université de Mascara, Mascara, Algerie ∥Faculty of Science and Technology, University of Siegen, Adolf-Reichwein Strasse 2, 57068 Siegen, Germany
ABSTRACT: The most efficient way to provide domestic lighting nowadays is by light-emitting diodes (LEDs) technology combined with phosphors shifting the blue and UV emission toward a desirable sunlight spectrum. A route in the quest for warm-white light goes toward the discovery and tuning of the lanthanide-based phosphors, a difficult task, in experimental and technical respects. A proper theoretical approach, which is also complicated at the conceptual level and in computing efforts, is however a profitable complement, offering valuable structure−property rationale as a guideline in the search of the best materials. The Eu2+-based systems are the prototypes for ideal phosphors, exhibiting a wide range of visible light emission. Using the ligandfield concepts in conjunction with density functional theory (DFT), conducted in nonroutine manner, we develop a nonempirical procedure to investigate the 4f7−4f65d1luminescence of Eu2+in the
environment of arbitrary ligands, applied here on the CsMgBr3:Eu2+-doped
material. Providing a salient methodology for the extraction of the relevant ligand field and related parameters from DFT calculations and encompassing the bottleneck of handling large matrices in a model Hamiltonian based on the whole set of 33 462 states, we obtained an excellent match with the experimental spectrum, from first-principles, without any fit or adjustment. This proves that the ligand field density functional theory methodology can be used in the assessment of new materials and rational property design.
■
INTRODUCTIONThe declaration, by the United Nations, of 2015 as the International Year of Light and Light-based Technologies1 as well as the recent award of the Nobel Prize for the invention of efficient blue light-emitting diodes (LED), which has enabled bright and energy-saving white light sources,2 put in sharp evidence the new impetus expected in such key domains, the energy saving turn of domestic lightening being one of the most immediate application goals. In the quest offinding phosphors able to alleviate the LED emission,3 bringing it as close as possible to sun daylight or the so-called warm-white light4−7 the systems based on lanthanide Eu2+-doped active centers are
among the best candidates.4−7 Besides, the luminescence of Eu2+ compounds is interesting in academic respects, being a
process essentially located at the atomic scale, due to the interconfiguration 4f7−4f65d1 transitions, which yield high
quantum efficiency even at room temperature.8−10 To complement the experimental efforts in the measurement the 4f7−4f65d1 excitation and the 4f65d1−4f7 emission spectra in various host materials, on the road of tuning proper compositions (host lattices and guest phosphors as doping),
the theoretical branches can be called as the source of rationale for the red lines of property design.
The theoretical accounts of the electronic structure and the properties of Eu2+considering its ground 4f7and itsfirst excited 4f65d1 configurations are relatively rare. This is because the electronic structure of two-open-shell 4f7and 4f65d1yields large matric interaction, namely, with 33 462 times 33 462 elements, generally prohibitive regarding computational cost. Therefore, the theoretical works are often tributary to standard ligandfield theories fed by empirical parameters.11−13Most of the studies, phenomenological or first principles, have used the advent of high symmetry focusing primarily to the octahedral cases.11−14 Nevertheless, the nature is more complicated than the preferred idealizations, the actual ongoing research4−7,15−21indicating the prevalence of systems where Eu2+ is in a lower coordination symmetry.
Herein we present a fully theoretical work based on density functional theory (DFT) enabling the investigation of the local structure of lanthanide-doped compounds, the calculation of
Published in ,QRUJDQLF&KHPLVWU\ ± which should be cited to refer to this work.
the electronic structure with respect to ligand field method-ologies, and the prediction of the optical effects of the 4f7−
4f65d1transitions of Eu2+embedded in the CsMgBr
3. This host
compound crystallizes in the hexagonal CsNiCl3structure type
with the space group P63/mmc (no. 194),22giving rise to an
intermediate symmetry case rather than the earlier octahedral situation. The lattice geometry of the CsMgBr3 host is
determined by means of DFT-based band structure calcu-lations. However, the doping of Eu2+ into CsMgBr
3 and
particularly the local distortion due to the presence of the Eu2+ impurity are explored by DFT geometry optimization operating with a selective cluster cut from the optimized CsMgBr3bulk.
On the basis of charge balance, a primary supposition is that the Eu2+impurity goes into the site formally occupied by the Mg2+
ion. Thus, the Eu2+ ion is coordinated by six bromide ligands
forming a D3d arrangement. The electronic structure of
(EuBr6)4− embedded in the CsMgBr3 host is discussed with
respect to ligand field analysis, showing the non-negligible influence of two Mg2+ions capping the (EuBr
6)4−octahedron.
The ligandfield potential in the Wybourne parametrization is calculated together with the Slater−Condon integrals and the spin−orbit coupling constants. They are used to calculate the multiplet energy levels and the oscillator strength of the electric dipole moment of the interconfiguration 4f7−4f6
5d1transitions. A priori, a special treatment of possible vibronic interaction in the 4f7configuration has permitted the observation of some line intensities belonging to the intraconfiguration 4f7−4f7
transitions, rendering therefore a realistic convoluted spectrum.
■
METHODOLOGYThe determination of the electronic structure of lanthanide-doped materials and the prediction of the optical properties are not trivial tasks. The standard ligand field models lack predictive power and undergoes parametric uncertainty at low symmetry, while customary computation methods, such as DFT, cannot be used in a routine manner for ligand field on lanthanide accounts. The ligand field density functional theory (LFDFT) algorithm23−30 consists of a customized conduct of nonempirical DFT calculations, extracting reliable parameters that can be used in further numeric experiments, relevant for the prediction in luminescent materials science.31These series of parameters, which have to be determined in order to analyze the problem of two-open-shell 4f and 5d electrons in lanthanide materials, are as follows.
(1) The gap parameterΔ(fd), which represents the energy shift of the multiplets of the 4fn−15d1 configuration with respect to those of the 4fn. It is obtained as the difference in the barycenter of the DFT energies associated with the 30 030 Slater-determinants arising from the 4f65d1manifold and those 3432 from the 4f7.23,32
(2) The Slater−Condon integrals Fk(ff), Fk(fd), and Gk(fd), which represent the static electron correlation within the 4fn and 4fn−15d1configurations. They are obtained from the radial wave functions Rnl of the 4f and 5d Kohn−Sham orbitals of the lanthanide ions.23,31
(3) The spin−orbit coupling constants ζ4fandζ5d, which represent the relativistic spin−orbit interaction in the 4f and 5d shells, also determined by means of the radial wave functions Rnlof the 4f and 5d Kohn−Sham orbitals of the lanthanide ions.23 (4) The Wybourne crystalfield parameters Bqk(f, f), Bqk(d, d), and
Bqk(f, d), which describe the interaction due to the presence of the ligands onto the electrons of the lanthanide center. They are deduced from the ligandfield energies and wave functions obtained from Kohn−Sham orbitals of restricted DFT calculations within the average of configuration (AOC) reference by placing evenly n− 1 electrons in the 4f orbitals and one electron in the 5d.33
The lattice geometry of CsMgBr3 is approached via periodical calculations by means of the VASP program package.34,35 The local density approximation (LDA) based on VWN36 and the generalized gradient approximation (GGA) formulized in PBE37are used for the exchange and correlation functional. The interaction between valence and core electrons is emulated with the projected augmented wave method.38,39 External as well as semicore states are included in the valence. A plane-wave basis set with a cutoff energy of 520 eV is used. Four k points were included in the direction a and b of the lattice, while 6 k points were comprised in the direction c.
The atomic positions were allowed to relax until all forces were smaller than 0.005 eV/Å. The local structure of the Eu2+impurity in CsMgBr3 is tackled via geometry optimization based on the cluster approach.40The optimized lattice structure of CsMgBr3 is used. An appropriate cluster is obtained as a selective cut along the 3-fold c axis of the unit cell of CsMgBr3(Figure 1a). A moiety containingfive units
of face-sharing (MgBr6)4−octahedra (Figure 1b) represents a perfect balance between the total number of atoms in the cluster (i.e., 35) and the charge of the edifice (i.e., + 4). This cluster is relaxed, keeping the electronic structure to have the 4f7and 4f65d1electron configurations of Eu2+, respectively. The positions of the atoms are relaxed, except those corresponding to Br(3), Mg(2), and their mirror images, as well as all Cs centers (Figure 1b), kept frozen to the optimized CsMgBr3 bulk, to mimic the presence of the lattice. The DFT calculations have been carried out by means of the Amsterdam Density Functional (ADF) program package (ADF2013.01),41−43which is one of the few DFT codes having the set of keywords facilitating the AOC calculations and Slater-determinant emulation needed by the LFDFT algorithm.23−33 The LDA VWN functional36 as well as the GGA PBE functional37are used for the geometry optimization. The molecular orbitals are expanded using triple-ζ plus two polarization Slater-type orbital (STO) functions (TZ2P+) for the Eu atom and triple-ζ plus one polarization STO function (TZP) for the Cs, Mg, and Br atoms. The LFDFT modeling of the electronic structure and the related optical properties of CsMgBr3:Eu2+is done using the hybrid B3LYP functional, in line with previous works.23,30−33
■
RESULTS AND DISCUSSIONLocal Structure of the Eu2+ Impurity. From the experimental perspective, the doping of lanthanide ions into solid state materials can be probed by different instrumental technics such as nuclear magnetic resonance (NMR),44 extended X-ray absorption fine structure (EXAFS),45,46 or electron paramagnetic resonance (EPR),47 which instead of giving a direct clue of the local geometry offers only data that can be corroborated to it. From the theoretical point of view,
Figure 1.Representation of the crystal structure of CsMgBr3(a) and a selective cut along the z axis showing five units of face-sharing (MgBr6)4−where the Eu2+dopant is placed into the position of the central Mg2+ion (b).
the classical way is the simulation by periodical crystal supercell structure23,31,32with band structure methodologies which are, however, not useful in the further advancement toward ligand field modeling.40Alternatively, one may consider a sufficiently
large cluster, carrying out geometry optimization by appropriate management of the variable versus frozen geometry coor-dinates. Using the same computational tools employed in the LFDFT account, this scheme has then inner consistency. In previous work32we probed that the cluster optimization leads to results similar to the bulk supercell approach. By means of band structure calculation, the optimized lattice parameters of the pristine CsMgBr3system are given in Table 1 compared
with the experimentally available X-ray diffraction data.48 We investigate the local structure around the Eu2+impurity, which is incorporated in the matrix of CsMgBr3(Figure 1). We use
the optimized structure of CsMgBr3obtained at the LDA and
GGA levels of theory (Table 1). Cesium magnesium bromide (CsMgBr3) crystallizes in the hexagonal P63/mmc space group
(no. 194).22,48 The divalent Eu2+ ions enter into the sites
formally occupied by Mg2+. They are coordinated by six bromide ligands within the D3dpoint group.
The geometry of the cluster represented inFigure 1b is then optimized. Theoretical results performed at two different levels of theory are collected inTable 2, compared with the reference structure of CsMgBr3 deduced from the experimental data.48
The cluster geometry optimization approach is used instead of the band structure algorithm in the investigation of the local structure of the Eu2+ impurity because this procedure yields appropriate structure with defined electron configuration.40 The 4f orbitals of Eu2+ split into a1u, a2u, a2u, eu, and eu
irreducible representations (irreps) of the D3d point group
(seeFigure 2). On the other hand, the 5d orbitals of Eu2+form the basis of a1g, eg, and eg irreps (see Figure 3). The ground
configuration geometry GC (Table 2) is reached distributing evenly seven electrons in the 4f orbitals of Eu2+. Then the
excited configuration geometry EC (Table 2) is attained by the self-consistency of an even distribution of six electrons in the 7-fold 4f orbitals, altogether with one electron placed in the a1g
component of the 5d orbitals. The LDA geometry optimization renders a closer match to the pristine CsMgBr3reference than
the GGA treatment (Table 2), i.e., it induces less perturbation in the pure lattice probably because the simpler approximation catches better the global bonding trends of the lattice.
Calculation of the Slater−Condon, spin−orbit cou-pling and ligand field parameters. The luminescence of CsMgBr3:Eu2+is crucially dependent on the local coordination
geometry of the Eu2+ dopant. Besides, a geometry change
occurs in the excited state 4f65d1(seeTable 2), leading to shifts
of lines in the emission spectrum as compared to the absorption (Stokes shift).49 Apparently intriguing, the excitation leads to a shortening of the Eu−Br(1) bond length (Table 2) because, contrary to the usual situation where orbital promotion increases the antibonding, the fact is that the occupation of the 5d orbitals strengthens the binding. This is because the 4f shell, having a radius smaller than the lanthanide ion itself, does not contribute to the chemical bonding, which is mostly ionic and partly due to donation effects into the 5d virtuals.40,50−52A surge of 5d electron population, by excitation, enhances the radial extension and bonding propensity of these functions, leading to shrinkage of the coordination radii.40 According to the LFDFT methodology,23,30 series of DFT calculations and subsequent analyses provide the ligand field one-electron parameters, the spin−orbit coupling constants (mainly one-electron character), and the related Slater− Condon integrals (two-electron, Coulomb, and exchange effects), further used in the full modeling of the complex Table 1. Optimized Lattice Parametersa, b, c (in
Angstroms) andα, β, γ (in degrees) of CsMgBr3Obtained at
the LDA and GGA DFT Levels of Theory, Together with the Experimental X-ray Diffraction Data (exp.)
LDA GGA exp.a
a 7.3342 7.8198 7.610 b 7.3342 7.8198 7.610 c 6.3240 6.6228 6.502 α 90.0 90.0 90.0 β 90.0 90.0 90.0 γ 120.0 120.0 120.0
aTaken from ref48.
Table 2. Optimized Local Structure Around the Eu2+ Impurity Embedded in CsMgBr3for the Ground 4f7(GC)
and Excited 4f65d1(EC) Electron Configurations of Eu2+ Obtained at the LDA and GGA DFT Levels of Theorya
GC EC
LDA GGA LDA GGA exp.b
Eu−Br(1) 2.8169 2.8607 2.7687 2.8189 2.6648 Eu−Mg(1) 3.3924 3.4161 3.3118 3.3287 3.2510 Eu−Br(2) 5.2453 5.3287 5.1867 5.2914 5.3140 Eu−Mg(2)c 6.3240 6.6230 6.3240 6.6230 6.5020 Eu−Br(3)c 8.1798 8.5480 8.1798 8.5480 8.3973 Mg(1)−Eu− Br(1) 51.46 52.15 53.99 54.56 52.41 Mg(1)−Eu− Br(2) 24.84 25.08 25.58 25.72 23.41 Mg(1)−Eu− Br(3)c 14.89 14.43 14.89 14.43 14.56
aDistances between the Eu2+dopant (in Angstroms) and next nearest neighbors and bond angles Mg−Eu−Br (in degrees) (Figure 1b).
bThe experimental data represent the crystal structure of CsMgBr
3 taken from ref48.cIn the cluster geometry optimization approach, the coordinates of Mg(2) and Br(3) species are kept frozen to the band structure data.
Figure 2. Representation of the energies of the 4f Kohn−Sham orbitals of Eu2+in CsMgBr
3obtained from the output of AOC-type calculation. The two diagrams represent the cases of geometries (Table 2) at ground and excited configurations (GC versus EC).
spectra.53−55 In order to meet the premises of ligand field theory, the DFT calculations must be performed under the so-called average of configurations (AOC) calculation proce-dure.23−33 It is a nonstandard handling that consists in imposing equal populations on the Kohn−Sham orbitals associated with the ligand field sequence, leading to a totally symmetric density. Then the orbital energies and eigenvectors are used to retrieve the ligand field matrix in atomic-like basis.23−33 The spin−orbit coupling, bringing the relativistic effects together with the Slater−Condon integrals, is derived from the radial functions of the 4f and 5d Kohn−Sham orbitals of Eu2+ averaged in the bromide environment for both
optimized structures inTable 2.23,31The calculated parameters are given inTable 3. The calculated ligandfield parameters33of the 4f and 5d orbitals of Eu2+are also shown inTable 3as a function of the GC or the EC geometry (Table 2). The ligand field energies of the 4f and 5d orbitals of Eu2+
are graphically represented inFigures 2and3, respectively, as a function of the GC and EC geometries (Table 2).
The ligandfield schemes are including the definite influence of the two Mg2+ ions capping the faces of the (EuBr6)4−
octahedron (Figure 1b). Their positive charges are exerting a stabilization trend against the electrons on the lanthanide, particularly for the orbitals pointing the lobes along the chain axis, i.e., fz3(Figure 2) and dz2(Figure 3) forming the basis of a2u and a1grepresentation in D3d, respectively. Indeed, in the
D3d symmetry of the (EuBr6)4−frame, the perturbation from
the two apical Mg2+ neighbors (Figure 1b) spans the representation of a1gand a2usymmetry.
The former leads to the stabilization of same symmetry component in the 5d diagram (Figure 3), while the latter interacts with the a2ufrom the 4f scheme (Figure 2). The ligand
field energy parameters are larger in the excited states (Table 3) because of the previously mentioned enhanced bonding capabilities. The slight radial expansion of the occupied 5d orbitals, in comparison to their status as virtuals, enhances their interaction with the chemical environment.
Angular Overlap Model (AOM) Analysis of the Ligand Field Potential. For a gain of chemical intuition, we rewrite the ligand field potential in terms of the angular overlap model.33The site symmetry of the Eu2+center is D3d, and the
ligand field splitting is mainly induced by the six equivalent Br(1) ligands as well as the two Mg(1) (Figure 1b). Therefore,
we can define the following set of parameters: eσ(f)Br−, eπ(f)Br−,
and eσ(f)Mg2+acting on the 4f orbitals and eσ(d)Br−, eπ(d)Br−, and
eσ(d)Mg2+ acting on the 5d ones. The definition of the AOM
parameters expresses the perturbation power of the environ-ment sites on theσ and π interaction channels. The eigenvalues of the ligand field matrix, suggested in Figures 2 and 3, respectively for 4f and 5d orbitals, are used to obtain AOM parameters by least mean squarefit. We calculated, for instance, in the case of the GC (Table 1) geometry the following parameters (in cm−1), eσ(f)Br−= 109, eπ(f)Br−= 92, eσ(f)Mg2+=
−26, eσ(d)Br−= 8639, eπ(d)Br−= 2900, and eσ(d)Mg2+=−1492,
the mean deviations with the DFT results (Figures 2 and3) being 7.39 and 0, respectively. The usual situation ranking of AOM parameters is eσ>|eπ| (occasionally with negative |eπ| for the case of back-donation effects) and |eλ(d)| ≫ |eλ( f)| with λ =
σ,π. Working with the nonstandard situation of accounting the perturbation from next neighbor cations, negative AOM parameters are obtained for the magnesium ions. This is perfectly reasonable considering that the positively charged sites must exert effects contrary to those emerging from the electron clouds of the ligands. One obtains a perfect mapping between the DFT results and the AOM formalism for the 5d ligandfield. However, in the 4f block, the small deviation can be assigned as due to higher order perturbations.33The splitting of the 4f orbitals obtained from the DFT calculation is much smaller if compared to the 5d orbital spacing. Then the DFT calculation takes into consideration many quantum effects that might not be fully classified by the simple σ and π interactions of the AOM formalism. Also, the AOM gives an intuitive
Figure 3. Representation of the energies of the 5d Kohn−Sham orbitals of Eu2+in CsMgBr3obtained from the output of AOC-type calculation. The two diagrams represent the cases of geometries (Table 2) at ground and excited configurations (GC versus EC).
Table 3. Calculated Slater−Condon Integrals, Spin−Orbit Coupling Constants, and Ligand Field Parameters (in cm−1) for CsMgBr3:Eu2+ Considering the Ground (GC) and
Excited (EC) Configurations Local Structures of the Eu2+
Impurity GC EC F2( ff) 388.47 388.47 F4( ff) 49.92 49.92 F6( ff) 5.30 5.30 G1( fd) 124.52 116.17 G3( fd) 11.00 8.64 G5( fd) 1.72 1.28 F2( fd) 87.01 76.12 F4( fd) 6.48 4.85 Δ(fd) 10 897.34 10 382.29 ζ4f 1246.50 1246.50 ζ5d 302.06 259.87 B02( f, f) −320.56 −723.85 B04( f, f) −475.07 −1194.36 B34( f, f) -i8.38 -i730.31 B−34( f, f) i8.38 i730.31 B06( f, f) −595.82 13.47 B36( f, f) -i291.33 -i152.72 B−36( f, f) i291.33 i152.72 B66( f, f) 88.92 127.23 B−66( f, f) 88.92 127.23 B02(d, d) 125.18 −3419.09 B12(d, d) −22 267.87 −20 525.90 B−12(d, d) −22 267.87 −20 525.90 B04(d, d) −24 470.57 −28 082.92 B14(d, d) 9090.82 8379.66 B−14(d, d) 9090.82 8379.66
http://doc.rero.ch
description of the ligand field potential being here able to indicate the contribution of the two Mg2+ with negative eσ
values.
Multiplet Structure and Absorption Spectra. The calculated multiplet energy levels arising from the 4f7 and
4f65d1 configurations of Eu2+ in CsMgBr3 are presented in
Figure 4 for the GC and EC geometries in Table 2. The 4f7
obtains 3432 energy levels constituting 1716 Kramers doublets, while the 4f65d1 leads to 30 030 states (15 015
Kramers-doublets). The 4f7ground state, 8S
7/2, with a very small
zero-field splitting is taken as the zero of the energy. The spectral energy levels of 4f7 (Figure 4, in red) are not significantly
affected by the GC versus EC geometry change, contradictory to those of 4f65d1 where a noticeable modification occurs
(Figure 4). The computed Stokes shift is 1748 cm−1, in the magnitude of the experimentally deduced value of 1920 cm−1.22 Considering the lowest margin of the 4f65d1 manifold, a 4f65d1−4f7emission at 19 885 cm−1in the green spectral region
(Figure 4) is predicted, also in agreement with the experimentally observed emission, i.e., 19 578 cm−122
The matrix element of the electric dipole moment operator d⃗αis constructed in terms of the one-electron wave functionψ
ψ ψ π
⟨ | ⃗ | ⟩ =μ αd v 34 ⟨Rn lμ μ| |r Rn lν ν⟩⟨Y Y Yn lμ μ| 1,α| n lν ν⟩ (1)
where in the right-hand side ofeq 1the term carrying the radial component is simple overlap integrals while the angular term is proportional with the Clebsh−Gordan coefficients.53,56
Consideringμ, ν = 4f and 5d,eq 1is a matrix with 12 by 12 elements, important for computation of the intensities of the transitions by distributing its elements over the whole manifold of the multielectron wave function. Considering the actual centrosymmetric D3d coordination of Eu2+, only the 7 by 5
elements corresponding to the off-diagonal ⟨ψ4f|d⃗α|ψ5d⟩ block
are nonzero. Thus, in D3dsymmetry, the 4f7→ 4f7transitions
do not carry intensity. However, vibronic coupling can easily encompass the Laporte restriction because the complete wave functionψ must include electronic and vibrational parts
ψμ=ψμorbital vibψμ (2)
To demonstrate this mechanism, we break the D3dsymmetry
of the complex. We have shown over the past two decades that multideterminantal DFT57−59 is adequate to describe the vibronic interaction. Therefore, we reoptimize the geometry (LDA structure) corresponding to GC (Table 2) without any symmetry constraint (C1), while previously we imposed the C3
axis of the mother lattice. A tiny distortion from the D3d
symmetry occurred with about 0.0072 Å average bond length change and a shallow energy stabilization of 241 cm−1. This affects by less than 5% the ligand field parameters, providing however a set of dipole moment integrals which (eq 1), taken in a relative sense, accounts for a weak intensity of the 4f7→ 4f7 transitions. This relaxation of the Born−Oppenheimer approximation is a convenient surrogate to describe the dynamic vibronic interaction
ψ ψ ψ ψ
⟨ 4forbital 4fvib| ⃗ |dα 4forbital 4fvib⟩ ≠0 (3)
that confers the intensity of the 4f7→ 4f7transitions.
The 4f7→ 4f65d1transitions are electric dipole allowed, the intensity matrix elements being roughly the same in the D3dand
C1geometries.
Using the ligandfield parameters, the two-electron integrals and spin−orbit coupling constants in Table 3 and the dipole matrix elements for the intensity part ineq 1, the prediction of the absorption spectrum is approached inFigure 5. One notes an exceptional match of the spectral profile of CsMgBr3:Eu2+22
with the first-principles simulation of energies and intensities. The dominant intensity comes from the dipole-allowed 4f7→ 4f65d1 transitions. Though small, the 4f7→ 4f7 part nicely accounts for the weak details visible as a distinct weak double peak placed at about 32 000 cm−1in the canyon between the large bands and the soft slope at about 37 000 cm−1, at the left margin of the second large band envelope (Figure 5).
The computed intensities of the predominant 4f7→ 4f65d1 transitions, represented by the height of the lines, are well matching the resolution details of the complex experimental spectrum.22Note that unless there is a global scaling, because of the arbitrary units in experimental data, nofit is implied in the representation of computed results.
The strong line intensity laying at 43 000 cm−1, while predicted, is not yet measured due to technical limits.22The virtue of LFDFT in retrieving realistic and detailed spectral data byfirst-principle route recommends the algorithm as a valuable complement in assessing various phosphor materials, along with experimental tests or by a priory property design, saving the tedious trial-and-error experimental combinatorial work.
Prediction of the Emission Spectrum. The prediction of the emission spectrum corresponding to the 4f65d1→ 4f7(8S)
transitions is rather delicate since the emission is the result of the relaxation from many states incorporating a humungous number of paths with nonradiative decay. By modest assumption, we can approximate the emission as the radiation from the lowest multiplet energy levels of the 4f65d1electron configuration (Figure 4) to the ground state of the 4f7, i.e., the 8
S7/2that is further split by very small zero-field splitting. Then
the electronic transition, which is electric dipole allowed, gives
Figure 4. Calculated multiplet energy levels (zero-phonon lines) arising from the ground 4f7 (in red) and excited 4f65d1 (in blue) configurations of Eu2+ in CsMgBr
3:Eu2+ obtained for the two geometries (LDA structures) representing the ground (GC) and excited (EC) configurations of Eu2+(Table 2). The energy range from 0 to 50 000 cm−1is selectively magnified showing the change in the 4f65d1energy levels of GC and EC. Some atomic spectral terms of the Eu2+in the 4f7configuration are given for clarity. All states, from spin octets to doublets, are included. (Inset) Full range spectrum.
rise to one zero-phonon line emission, which is further coupled to vibrations. We use the so-called Huang−Rhys factor gk
59−64
to measure the strength of the linear electron−phonon coupling. Since the transition path, which is used for the transformation of the EC geometry to the GC one (Table 2), conserves the D3d symmetry, the Huang−Rhys factor are
mainly associated with the totally symmetric vibrational mode a1g. It is straightforward to consider the neutral cluster
(EuBr6Mg2), whose coordinates are taken from Table 2, for
the two stationary points GC and EC representing the structure of the Eu2+impurity at the ground and excited configurations of Eu2+ in CsMgBr3, respectively. The analytical harmonic
frequencies and normal modes at the two stationary points are determined. They are presented inTable 4, where there are 21 normal modes belonging to the neutral cluster which form the basis of the following irreps: 3 a1g, 3 eg, a1u, 3 a2u, and 4 eu
(Table 4). The correlation between the normal modes of the GC and EC are constructed by means of the orthonormal Dushinsky matrix.59,65 The displacement vector is the di ffer-ence between EC and GC stationary points, and k expresses the contribution along the mass-weighted GC normal coordinates to the displacement vector. Ek represents the energy
contribution of the normal modes on the relaxation of the EC stationary point to the GC one within the 4f7 electron configuration of Eu2+.59
They are also represented in Table 4. The Huang−Rhys factor gkis related to k following ref63.
The calculated values of the Huang−Rhys factor are also collected inTable 4, where it is seen that the contribution of
the totally symmetric a1gvibrational modes are the biggest, i.e.,
they are mainly responsible for the electron−phonon coupling. A graphical representation of their motion is given inFigure 6.
To simulate the vibrational progression, we obtain the Franck−Condon factors using the two-dimensional array method in ref 64. We consider 1 vibrational quantum (ν = 0) from the EC stationary point and 21 vibrational quanta (ν′ = 0, 1, ..., 20) from the GC stationary point. The Franck− Condon factors are then calculated for every permutation up to 21 quanta over the vibrational modes. It is necessary in order to get all Franck−Condon factors of the EC stationary point with respect to each three a1gvibrational state (Figure 6) of the GC
to sum to one. One obtains a qualitative agreement between the calculated and the experimental emission profiles (Figure 7).
■
CONCLUSIONIn this work, we outline the methodology for realistic analysis and first-principles prediction of the optical properties useful for obtaining phosphors, ameliorating the emission of LED devices by shifting the blue and UV predominance realizing
Figure 5. (Right) Calculated electric dipole oscillator strength (i.e., zero phonon lines) obtained for the transitions 4f7(8S)→ 4f65d1(in violet) and 4f7 (8S) → 4f7 (in magenta) of Eu2+ in CsMgBr
3:Eu2+ together with the experimental excitation spectra (in black) taken from ref22. The multiplet energy levels corresponding to the 4f7(in red) and 4f65d1 (in blue) are presented on the bottom line. The green curve represents a superimposition of a Gaussian band with a width of 125 and 350 cm−1on the zero phonon lines corresponding to the 4f7 (8S)→ 4f7and 4f7(8S)→ 4f65d1transitions, respectively. (Left-upper corner) Mechanisms of phosphor action, converting high-energy absorbed quanta to lower energy emission via arrays of nonradiative decay along the many states of the 4f65d1manifold comprised in the first band. (Left-side inset) Magnified sequence containing the 4f7(8S) → 4f7weak intensity bands.
Table 4. Predicted Vibrational Energiesℏω (in cm−1) Corresponding to the GC and EC Stationary Points, The Displacement Vectork (in (amu1/2·Å)), the Energy ContributionEk(in cm−1), and the Huang−Rhys Factor gk
(dimensionless) ℏω vib. GC EC k Ek gk eu 38.12 42.55 0.0000 0.00 0.00 a1u 41.78 31.90 0.0000 0.00 0.00 a1g 72.42 82.93 −3.0843 647.39 10.22 eu 78.15 63.23 0.0000 0.00 0.00 eg 83.39 62.44 −0.0017 0.00 0.00 a2u 94.68 91.85 0.0000 0.00 0.00 eg 137.70 147.54 −0.0013 0.00 0.00 a1g 161.03 169.42 −0.4170 73.27 0.42 eu 166.62 171.18 0.0000 0.00 0.00 a2u 185.20 206.97 0.0000 0.00 0.00 eu 214.64 214.82 0.0000 0.00 0.00 eg 215.52 183.06 −0.0007 0.00 0.00 a1g 267.28 262.71 −1.0481 137.61 4.35 a2u 283.27 239.16 0.0000 0.00 0.00
Figure 6.Representation of the three totally symmetric a1gvibrations of (EuBr6Mg2) that is responsible of the electron phonon coupling in the 4f65d1−4f7transitions of Eu2+ in CsMgBr
3. Color code: Eu2+in violet, Br−in red, and Mg2+in yellow.
warm-white light. The LFDFT method consists of nonroutine DFT calculations setting numeric experiments defining the ligand field and related parameters (Slater−Condon integrals and spin−orbit coupling). Without the designed handling of spin and orbital population, the DFT in itself does not directly provide information consistent with the ligandfield frame. The ligandfield is a necessary ancillary tool, producing simulations of spectra, extending therefore the power of DFT, from where the parameters originate, beyond its nominal reach. The method was applied to CsMgBr3 doped with Eu2+, where a
LFDFT modeling of (EuBr6)4− octahedra belonging to an
extended cluster mimicking well the local and long-range environment has been performed. The considered problem has a high level of technical difficulty in the ligand field modeling part, implying a nontrivial case of two-open-shell ligand field and a large dimension of Hamiltonian matrix, in general prohibitive, encompassed with special algorithmic steps of parallel computing. The match of experimental computed profiles to the experimental spectrum, in both major pattern and minor details, without adjustment or fit proves that the developed theoretical method is a reliable complement to experimental analysis.
■
AUTHOR INFORMATION Corresponding Authors *E-mail:harry.ra@hotmail.com. *E-mail:cfanica@yahoo.com. *E-mail:wurland@arcor.de. *E-mail:claude.daul@unifr.ch. NotesThe authors declare no competingfinancial interest.
■
ACKNOWLEDGMENTSThe Swiss National Science Foundation (SNF) and the Swiss State Secretariat for Innovation and Research supported this work. Funds from the UEFISCDI Romania research grant PCE 14/2013 are also acknowledged.
■
REFERENCES(1) 2015 International year of light and light-based technologies, availablehttp://www.light2015.org/Home.html[May 1, 2015].
(2) Heber, J. Nat. Phys. 2014, 10, 791.
(3) Nakamura, S.; Fasol, G. The blue Laser Diode; Springer: Berlin, 1997.
(4) Jüstel, T.; Nikol, H.; Ronda, C. Angew. Chem., Int. Ed. 1998, 37, 3084.
(5) Höppe, H. A. Angew. Chem., Int. Ed. 2009, 48, 3572. (6) Bünzli, J.-C. G.; Piguet, C. Chem. Soc. Rev. 2005, 34, 1048. (7) Eliseeva, S. V.; Bünzli, J.-C. G. Chem. Soc. Rev. 2010, 39, 189. (8) Tyner, C. E.; Drickamer, H. G. J. Chem. Phys. 1977, 67, 4116. (9) Kim, J. S.; Park, Y. H.; Kim, S. M.; Choi, J. C.; Park, H. L. Solid State Commun. 2005, 133, 445.
(10) Dorenbos, P. J. Phys.: Condens. Matter 2005, 17, 8103. (11) Burdick, G. W.; Burdick, A.; Deev, V.; Duan, C.-K.; Reid, M. F. J. Lumin. 2006, 118, 205.
(12) Pan, Z.; Ning, L.; Cheng, B.-M.; Tanner, P. A. Chem. Phys. Lett. 2006, 428, 78.
(13) Garcia-Fuente, A.; Cimpoesu, F.; Ramanantoanina, H.; Herden, B.; Daul, C.; Suta, M.; Wickleder, C.; Urland, W. Chem. Phys. Lett. 2015, 622, 120.
(14) Aiga, F.; Hiramatsu, R.; Ishida, K. J. Lumin. 2014, 145, 951. (15) Marchuk, A.; Schnick, W. Angew. Chem., Int. Ed. 2015, 54, 2383. (16) Piao, X.; Machida, K.-I.; Horikawa, T.; Hanzawa, H.; Shimomura, Y.; Kijima, N. Chem. Mater. 2007, 19, 4592.
(17) Ruan, J.; Xie, R.-J.; Hirosaki, N.; Takeda, T. J. Am. Ceram. Soc. 2011, 94, 536.
(18) Xia, Z.; Zhang, Y.; Molokeev, M. S.; Atuchin, V. V. J. Phys. Chem. C 2013, 117, 20847.
(19) Bachmann, V.; Ronda, C.; Oeckler, O.; Schnick, W.; Meijerink, A. Chem. Mater. 2009, 21, 316.
(20) Krings, M.; Montana, G.; Dronskowski, R.; Wickleder, C. Chem. Mater. 2011, 23, 1694.
(21) Bachmann, V.; Jüstel, T.; Meijerink, A.; Ronda, C.; Schmidt, P. J. J. Lumin. 2006, 121, 441.
(22) Suta, M.; Larsen, P.; Lavoie-Cardinal, F.; Wickleder, C. J. Lumin. 2014, 149, 35.
(23) Ramanantoanina, H.; Sahnoun, M.; Barbiero, A.; Ferbinteanu, M.; Cimpoesu, F. Phys. Chem. Chem. Phys. 2015, 17, 18547.
(24) Atanasov, M.; Daul, C. A.; Rauzy, C. Struct. Bonding (Berlin) 2004, 106, 97.
(25) Atanasov, M.; Comba, P.; Daul, C. J. Phys. Chem. A 2006, 110, 13332.
(26) Atanasov, M.; Daul, C.; Gudel, H. U.; Wesolowski, T. A.; Zbiri, M. Inorg. Chem. 2005, 44, 2954.
(27) Atanasov, M.; Comba, P.; Daul, C. Inorg. Chem. 2008, 47, 2449. (28) Atanansov, M.; Daul, C. Chimia 2005, 59, 504.
(29) Atanansov, M.; Daul, C. C. R. Chim. 2005, 8, 1421.
(30) Ramanantoanina, H.; Urland, W.; Cimpoesu, F.; Daul, C. Phys. Chem. Chem. Phys. 2013, 15, 13902.
(31) Ramanantoanina, H.; Urland, W.; Garcia-Fuente, A.; Cimpoesu, F.; Daul, C. Phys. Chem. Chem. Phys. 2014, 16, 14625.
(32) Ramanantoanina, H.; Urland, W.; Garcia-Fuente, A.; Cimpoesu, F.; Daul, C. Chem. Phys. Lett. 2013, 588, 260.
(33) Ramanantoanina, H.; Urland, W.; Cimpoesu, F.; Daul, C. Phys. Chem. Chem. Phys. 2014, 16, 12282.
(34) Kresse, G.; Hafner. Phys. Rev. B: Condens. Matter Mater. Phys. 1993, 47, 558.
(35) Kresse, G.; Furthmüller, J. Phys. Rev. B: Condens. Matter Mater. Phys. 1996, 54, 11169.
(36) Vosko, S. H.; Wilk, L.; Nussair, M. Can. J. Phys. 1980, 58, 1200. (37) Perdew, J. P.; Burke, K.; Ernzerhof, M. Phys. Rev. Lett. 1996, 77, 3865.
(38) Blöchl, P. E. Phys. Rev. B: Condens. Matter Mater. Phys. 1994, 50, 17953.
(39) Kreese, G.; Joubert, D. Phys. Rev. B: Condens. Matter Mater. Phys. 1999, 59, 1758.
Figure 7. Calculated Franck−Condon factors of the EC stationary point with respect to the a1gno. 1 (in red), a1gno. 2 (in yellow), and a1gno. 3 (in blue) vibrational states (Figure 6) of the GC stationary point. The green curve represents a superimposition of a Gaussian band with a width of 350 cm−1on the Franck−Condon factors. The experimental observed emission spectrum22for the 4f65d1→ 4f7(8S) transitions is given in black.
(40) Ramanantoanina, H.; Urland, W.; Herden, B.; Cimpoesu, B.; Daul, C. Phys. Chem. Chem. Phys. 2015, 17, 9116.
(41) te Velde, G.; Bickelhaupt, F. M.; van Gisbergen, S. J. A.; Guerra, C. F.; Baerends, E. J.; Snijders, J. G.; Ziegler, T. J. Comput. Chem. 2001, 22, 931.
(42) Guerra, C. F.; Snijders, J. G.; te Velde, G.; Baerends, E. J. Theor. Chem. Acc. 1998, 99, 391.
(43) Baerends, E. J.; Ziegler, T.; Autschbach, J.; Bashford, D.; Berces, A.; Bickelhaupt, F. M.; Bo, C.; Boerrigter, P. M. ; Cavallo, L.; Chong, D. P.; Deng, L.; Dickson, R. M.; Ellis, D. E.; van Faassen, M.; Fan, L.; Fischer, T. H.; Guerra, C. F.; Ghysels, A.; Giammona, A.; van Gisbergen, S. J. A.; Göẗz, A. W.; Groeneveld, J. A.; Gritsenko, O. V.; Grüning, M.; Gusarov, S.; Harris, F. E.; van den Hoek, P.; Jacob, C. R.; Jacobsen, H.; Jensen, L.; Kaminski, J. W.; van Kessel, G.; Koostra, F.; Kovalenko, A.; Krykunov, M. V.; van Lenthe, E.; McCormack, D. A.; Michalak, A.; Mitoraj, M.; Neugebauer, J.; Nicu, V. P. ; Noodleman, L.; Osinga, V.P.; Patchkovskii, S.; Philipsen, P. H. T.; Post, D.; Pye, C. C.; Ravenek, W.; Rodriguez, J. I.; Ros, P.; Shipper, P. R. T.; Schreckenbach, G.; Seldenthuis, J. S.; Seth, M.; Snijders, J. G.; Sola, M.; Swart, M.; Swerhone, D.; te Velde, G.; Vernooijs, P.; Versluis, L.; Vissher, L.; Visser, O.; Wang, F.; Wesolowski, T.; van Wezenbeek, E. M.; Wiesenekker, G.; Wolff, S. K.; Woo, T. K.; Yarkolev, A. L. ADF2013.01, available athttp://www.scm.com.
(44) Walstedt, R. E.; Walker, L. R. Phys. Rev. B 1974, 9, 4857. (45) O’Day, P. A.; Rehr, J. J.; Zabinsky, S. I.; Brown, G. E. J. Am. Chem. Soc. 1994, 116, 2938.
(46) Ghigna, P.; Carollo, A.; Flor, G.; Malavasi, L.; Peruga, G. S. J. Phys. Chem. B 2005, 109, 4365.
(47) Padlyak, B. V.; Grinberg, M.; Lukasiewicz, T.; Kisielewski, J.; Swirkowicz, M. J. Alloys Compd. 2003, 361, 6.
(48) McPherson, G. L.; McPherson, A. M.; Atwood, J. L. J. Phys. Chem. Solids 1980, 41, 495.
(49) Dorenbos, P. J. Lumin. 2000, 91, 155.
(50) Cimpoesu, F.; Dragoe, N.; Ramanantoanina, H.; Urland, W.; Daul, C. Phys. Chem. Chem. Phys. 2014, 16, 11337.
(51) Clark, D. L.; Gordon, J. C.; Hay, P. J.; Poli, R. Organometallics 2005, 24, 5747.
(52) Neidig, M. L.; Clark, D. L.; Martin, R. L. Coord. Chem. Rev. 2013, 257, 394.
(53) Griffith, J. S. The theory of Transition Metal Ions; Cambridge University Press: Cambridge, 1961.
(54) Figgis, B. N.; Hitchman, M. A. Ligand field theory and its applications; Wiley-VCH: New York, 2000.
(55) Gerloch, M.; Harding, J. H.; woodley, R. G. Struct. Bonding (Berlin) 1981, 46, 1.
(56) Cowan, R. D. The theory of atomic structure and spectra; University of California Press: Berkeley, 1981.
(57) Bruyndonckx, R.; Daul, C.; Manoharan, T. Inorg. Chem. 1997, 36, 4251.
(58) Ramanantoanina, H.; Gruden-Pavlovic, M.; Zlatar, M.; Daul, C. Int. J. Quantum Chem. 2013, 113, 802.
(59) Ramanantoanina, H.; Zlatar, M.; Garcia-Fernandez, P.; Daul, C.; Gruden-Pavlovic, M. Phys. Chem. Chem. Phys. 2013, 15, 1252.
(60) Wilson, E. B., Jr. An Introduction to Scientific Research; McGraw-Hill International Book Co., Inc.: New York, 1952.
(61) Balhausen, C. J. Molecular Electronic Structure of Transition Metal Complexes; McGraw-Hill International Book Co., Inc.: New York, 1979.
(62) Sharp, T. E.; Rosenstock, H. M. J. Chem. Phys. 1964, 41, 3453. (63) Moreno, M.; Barriuso, M. T.; Aramburu, J. A. J. Phys.: Condens. Matter 1992, 4, 9481.
(64) Ruhoff, P. T.; Ratner, M. A. Int. J. Quantum Chem. 2000, 77, 383.
(65) Duschinsky, F. Acta Physicochim. URSS 1937, 7, 551−566.
